(Preview) Deep Research and OpenAI's Business Model, A Counterpoint on White Collar Concerns, The Future of Cognition and Companionship
14:11 Share

Hello and welcome to a free preview of Sharp Tech. Yeah, it seems like a natural follow-up question when you use deep research and then think about how valuable private information is going to be going forward. It's like, how did deals get made to integrate some of this stuff or make it easy for users to integrate some of this stuff? And I'll be curious to see how they tackle it going forward.

Right. And there's also this real tension between the, you know, just the cost of doing this and serving it. You're feeding it all these PDFs. You're doing all these sorts of things. Does this actually make sense on a subscription basis? And some people drastically overuse it relative to your cost basis.

Does that is that made up for by people like me that are maybe under using it but are still paying $200 a month or is this or do you buy on a one off basis and it's worth it because you bought it? I thought one of the interesting things talking to Sam about this was his I thought pretty strong.

I wouldn't say strident, but of course I'm open to anything, blah, blah, blah, but strong dismissal, I think, of an advertising-based model. I would say strident. I was pretty shocked by... I mean, he didn't outright dismiss the possibility, but he came back to it a couple times. He just does not really believe in ads, and they're trying to prioritize other business models right now. Yeah, and I think that's a mistake. I mean, I think it's missing their opportunity,

And as I think I said to him, the way you get – or maybe this was afterwards. We sort of talked a little bit after. I'm like any ad product you launch is going to suck. The question is does it suck in 2025 or does it suck in 2028? Because the way to get good ones is to just get out there and do it and to iterate. I wrote an article ages ago about Apple and culture and one of the challenges they had is the nature of web services online.

is it's the exact opposite of a phone. The whole challenge with building a phone is you make the phone once and you can't

change it. Right now, that's not totally right. They do make adjustments like they get the initial feedback of people bringing their phones back. What's wrong with them? XYZ. They will make small adjustments in the production line, but that will affect phones weeks or months down the road, right? Like they have to deal with the ones that are there and that forces a certain sort of mindset and culture, which is you, you

you know, measure twice, cut once because you only get to cut once. Right. And I think this is one of the reasons why Apple traditionally sucked at web services. You go back to the mobile me thing. They, it was this beautiful interface and it did all this sort of stuff.

And then it fell apart once real people started using it because the nature of the web is you can't do that. There's so many variables. There's so many things you have to be able to fail gracefully. And you fail at multiple steps and you can recover and then it's super iterative. You can be updating and push it right away. But it's also just the nature of it is you –

with software in general, there's always bugs in software. Every time you ship, there's bugs. It's a decision our...

are the bugs that we know are there acceptable and knowing there's also a lot of bugs we don't know are there that are going to show up when they get there? And this is a hard thing to get right. Like, we had a pretty rough passport update in January where basically we got this balance wrong. And, you know, we're doing things internally to revisit this and, you know, stay tuned. There's going to be like a beta program and things along those lines that we're going to try to make sure we don't have a similar situation

episode again, but it's hard. It's a hard problem and it's very different than shipping a phone. And the nature of being a web service and something like advertising is like that. You're only going to get good at it by doing it. You can't be precious about it. And there was, I feel, a degree of preciousness that was coming across

from Sam and about their product, which didn't surprise me. It's my expectations and assumptions about open AI kind of generally, but I do think it's a big internal obstacle to them achieving what I think they can achieve, which is the hardest thing to do, which is a consumer aggregator destination site. Like I told him, they're the first one since Facebook. That's how rare this is. It's like an every 15 year sort of event and,

And you're only going to get good by doing it. And even if it sucks to start. Yeah. And the other thing that resonated as I was reading the interview was you talking about your time writing Stratechery. There has not been a company to emerge since you've been writing Stratechery. Like all the dominant consumer tech companies were in place at that.

Yeah.

But how big is that market really? How many high agency people are out there that are willing to understand this tool, pay $200 a month and take advantage of it? And, you know, I think that is a pretty small market at the end of the day relative to what's possible with a billion people otherwise. Yeah, well, I mean, I think this just applies to subscription products in general.

I know with Stratechery, my market is limited. Like even though I keep the price relatively low because I want to have a larger market than I might otherwise. But the fact of the matter is your own, like you're, I'm just never going to be the size and scale of anything that's free. And that's even before you get to the whole aggregator concept and capturing the whole world and all those sorts of pieces. And I think, yeah, limiting themselves to a subscription-based business model is

That's a nice thing for an anthropic to do. I think it's a dereliction of potential for open AI to do. They should be thinking bigger. And the reality is when it comes to consumer tech companies, bigger means advertising. That's just the way it is. And the thing about advertising is it's not...

People have such a bias against it. I've defended advertising for ages and ages. And again, there is an irony where I don't do advertising, but that's because I'm not big. Like for someone of my scale, subscriptions are great. When you're big, the magic of advertising is the whole world is your addressable market because it's free.

And you get to raise prices on customers because it's the advertisers that are paying more, not the customers. And that's why it's inevitable that a large consumer tech company or at least a web services company is going to be ad supported. That's the way you grow revenue in the long run. And even if you want to have a middle ground, like say a Netflix, like the –

Netflix's growth has been so large because they added the ad product. In part, they were saturated on people willing to pay $15 a month or $18 a month or whatever what it's up to now. And so that lets them charge more. It opens up that funnel at the top to move them down. And yeah, I think it's a mistake. And I think the longer they wait, the...

it'll be worse for the business. And the more likely it is, they just missed the opportunity as a whole and someone comes along. And the other thing is people that are high agency that use these tools and will pay money

Believe me, every Anthropic release, I'm checking it out. I'm testing it. I already talked about it. I use Grok much more than OpenAI now. Like, yeah, you're fighting for the hardest customers to keep. Well, you talk about a bias against advertising. One note is I am insulated from the state of advertising on the open internet these days because I'm mostly just reading Substack and reading Stratechery and subscriptions. And...

And, you know, Financial Times, Bloomberg, whatnot. But every year around the NCAA tournament, I'll seek out information from the open web. Make your pool picks. Exactly. And like, man, oh, man. Who do you have winning, Andrew?

I have Auburn winning. Oh, see, I knew it. There is one clear choice this year based on the numbers and the metrics. One team is top five in defensive efficiency and offensive efficiency. It is Duke basketball. There's no way.

I chose Auburn because Bruce Pearl has always struck me as like a scumbag extraordinaire college basketball coach. And college basketball rewards the scumbags. So I'm pulling for the Auburn Tigers. But the point is, you go to CBS Sports, you go to Sports Illustrated and

And the experience has just become unbelievably degraded and depressing. So I understand why advertising may make people's skin crawl. Yeah, but you know which ads don't suck? Facebook. That's right. Facebook ads don't suck. Instagram ads don't suck. I haven't been to Facebook in years. But Instagram, they do it elegantly. Yes, they do. Now, it's not clear how you get...

Those quality of ads in an AI experience, right? Like, you know, the start out, you're just having like display hats, right? And you have great targeting because people are typing in their entire like world into your, your little box, but yeah, it's going to suck and that's painful. But the way you get to not sucking is by doing. And, uh, so, uh,

Yeah, we'll see whether they do it. I do have to make an admission. I've been meaning to mention this, I think, for a long time. Okay. Early on in Stratechery, I was very critical of people that like ad blockers and all that sort of thing. And I think I might have used – it was an exponent. I call these people entitled, which really set a lot of people off. I bet. Yeah.

And it's like, hey, if you want the access, like, that's how you pay for it is the ads, right? Just don't read it if you don't like the ads. Oh, my God. Holier than thou, Ben. I missed that era. Yeah. So at some point, the web got so bad. I use the web a lot. Mm.

I had to do, I think it was Chrome, like the one you use with like uBlock Origin or something on those lines, which now is like deprecated because Google updated Chrome, I'm sure for very valid reasons, just happened to kill the best ad blocker. And so like for a day, I didn't have an ad blocker and it was...

horrific and i just needed to tell my audience who's been here a long time i'm now entitled i'm sorry i called you that yeah it was i really i just need to come clean it's it's uh comes down from the mountaintop hand in hand signing up for an ad blocker uh it's it's necessary and i'm a bad person but

I hate to see it. We do what you got to do. Yeah. Well, it's bad out there. You know, you need an ad blocker. It's so rough. All right. Well, to keep it moving, one more question on deep research. Mason says, on Sharp Tech a few weeks ago, Andrew gave an example from when he was a young lawyer and a partner asked him to put together a memo of legal research. It took Andrew about six hours to do the research and write the memo. Now, with deep research from OpenAI, Andrew said it would take about 20 minutes.

The concern highlighted by Andrew was the potential displacement of junior lawyers, consultants, and other white-collar-type positions.

I myself work in equity research for a large bank. I would characterize my job as similarly at risk with new AI tools like deep research. On the other hand, I was recently part of a discussion with a much older, seasoned equity research analyst who said in jest, what do associates even do these days? Back in the 80s and 90s when I was coming up in the business, I would work 15-hour days compiling spreadsheets and charts by hand.

Well, 30 to 40 years later, the equity research profession still exists and likely employs even more people than it did back then.

So what gives? I'm not spending 15 hours a day making charts and compiling data because that takes me minutes using Microsoft Excel and other tools. It's not that our jobs were rendered irrelevant because of better technology. Rather, the nature of the work changed. Research analysis and associates took the time freed up from not having to hand calculate spreadsheets and instead do other more value added research.

What do you guys think? My thesis is that this time isn't necessarily different. Game-changing technological breakthroughs happen in every generation, and I don't think AI will lead to mass unemployment, but rather a redistribution of knowledge workers' time and attention to more value-added tasks and jobs.

So I wanted to include this because we got a few variations of this response to the Deep Research podcast earlier this year. And it's a refrain that emerges pretty consistently in conversations around AI and its impact on the economy. So what do you think of that line of reasoning, Ben?

All right, and that is the end of the free preview. If you'd like to hear more from Ben and I, there are links to subscribe in the show notes, or you can also go to sharptech.fm. Either option will get you access to a personalized feed that has all the shows we do every week, plus lots more great content from Stratechery and the Stratechery Plus bundle. Check it out, and if you've got feedback, please email us at email at sharptech.fm.